1. Field of the Disclosure
Embodiments of the invention relate generally to image sensor testing methods and related testing apparatuses. More particularly, embodiments of the invention relate to automated test methods for image sensors and related test apparatuses providing improved testing uniformity.
This application claims the priority of Korean Patent Application Nos. 2006-0002296, filed Jan. 9, 2006, and 2006-0022314, filed Mar. 9, 2006, the disclosures of which are hereby incorporated by reference in their entirety.
2. Description of the Related Art
Improved semiconductor fabrication technology has resulted in semiconductor image sensors characterized by smaller size, high quality performance, and lower manufacturing costs. As a result, image sensors are now used in a wide variety of commercial devices, such as digital cameras, camcorders, printers, scanners, and certain cellular telephones. In general operation, image sensors are adapted to capture optical energy (e.g., visible light), convert the optical energy into coherent electrical signals, and subsequently process the electrical signals into digital data that may be easily stored, transferred, and manipulated. The end result of this remarkable sequence of processes is the generation of image data that may be visually displayed or recorded in a number of different forms using conventionally available digital media.
The two most widely used types of semiconductor image sensors are the charge coupled device (CCD) and the CMOS image sensor (CIS). CCD sensors generally provide higher performance than CIS sensors. CCD sensors operate with comparatively lower noise and higher device uniformity. However, CIS sensors operate with comparatively lower power consumption and at higher operating speeds. The lower power consumption and higher operating speed provided by CIS sensors make CIS sensors the image sensor of choice for many portable electronic devices. That is, for many portable electronic devices constrained in their design and operating characteristics by a limited battery life, the reduced power consumption is an acceptable trade-off with the inferior performance of CIS sensors relative to CCD sensors. Cellular telephones having an integrated camera are one example of this type of portable electronic device.
This current design preference related to portable electronic devises raises some difficult challenges in the context of the constituent image sensors. For example, given the inherent weaknesses associated with CIS sensors (e.g., high noise and less uniformity between individual sensor operations), the quality control and testing processes within the lengthy sequence of fabrication processes necessary to manufacture CIS sensors become increasingly important.
Regardless of whether CCD or CIS technology is used, most image sensor manufacturers begin quality control testing at the wafer level to avoid the expense of continued fabrication of failed image sensor elements, and to reduce the possibility of producing non-functional or substandard products. The term “wafer level” refers to any fabrication process, including testing and quality control processes, performed on a wafer substrate containing many individual semiconductor image sensors. (See, Figure (FIG.) 1). Wafer level testing is performed before individual devices are cut from the wafer and packaged or connected to other host device components.
Given the optical-to-electrical conversion functionality provided by image sensors, it is not surprising that optical testing (e.g., a test associated with the optical illumination of the image sensor) is an important part of the overall quality control process. A great deal of optical testing is performed at the wafer level. Such testing directly implicates the quality of the optical test equipment and, more particularly, the quality of the illumination source within the optical test equipment. Indeed, the reliability of the optical test equipment should at all times exceed the performance specifications of the image sensors being tested. A consequence of this conclusion requires that the illumination source within optical test equipment be precisely controlled in its characteristics (e.g., intensity, wavelength, etc.), highly uniform in its application across a wafer, and very reliable. Without such qualities in an illumination source, it is impossible to distinguish performance variations between individual image sensors formed on the wafer.
As currently used in the fabrication of image sensors, conventional illumination sources are manually calibrated according to a periodic maintenance schedule (e.g., every 1000 to 2000 hours of operation). Such an approach requires interruptions of the fabrication process, as test equipment is taken off-line for maintenance. Further, the manual calibration process is subject to qualitative variations due to its inherent “human factors” (e.g., variations in the training of technicians, etc.).
Additionally, the mass production capabilities provided by modern semiconductor fabrication facilities mandates the use of multiple optical testing stations, each having an illumination source. Ideally, the illumination source used in each fabrication line should produce an identical output so that image sensor testing across multiple fabrication lines is consistent. Unfortunately, this is almost never the case. As presently instituted, each illumination source is merely tested (or compared) to a “standard” illumination source, and this process is inherently variable in its outcome.
The difficulty of consistently providing uniform, high quality illumination sources is compounded by the fact that the performance of all illumination sources tends to degrade over time. As a result, illumination source manufacturers provide expected performance profiles that characterize “typical” illumination performance over time. These performance profiles suggest voltage bias adjustments or operating voltage offset values for an illumination source during different periods of its life span. However, such performance profiles and the corresponding voltage offsets are merely “average” adjustments defined in relation to modeled outcomes. They do not accurately take into consideration the actual performance of individual illumination sources.
All of the foregoing results in an unacceptably high level of variation in the performance quality of illumination sources. Absent a reliable and clearly defined optical reference signal (e.g., uniform optical energy having a well controlled intensity) it is not possible to accurately characterize the performance of individual image sensors formed on a wafer. More critically, it is well understood that semiconductor fabrication processes frequently produce variable results across the surface area of a wafer (e.g., material layer characteristics on edge portions of the wafer verses center portions of the wafer). Such fabrication process variations must be identified, carefully quantified, and controlled in order to improve image sensor yield. It is nearly impossible to identify such process variations during wafer level optical testing of image sensors if the characteristics of the illumination source in the optical test equipment also vary across the surface of the wafer.